Original Investigation | Oncology

Role of Diet in Colorectal Cancer IncidenceUmbrella Review of Meta-analyses of Prospective Observational StudiesSajesh K. Veettil, PhD; Tse Yee Wong, B Pharm; Yee Shen Loo, B Pharm; Mary C. Playdon, PhD; Nai Ming Lai, MRCPCH;Edward L. Giovannucci, MD, ScD; Nathorn Chaiyakunapruk, PharmD, PhD

Abstract

IMPORTANCE Several meta-analyses have summarized evidence for the association betweendietary factors and the incidence of colorectal cancer (CRC). However, to date, there has been littlesynthesis of the strength, precision, and quality of this evidence in aggregate.

OBJECTIVE To grade the evidence from published meta-analyses of prospective observationalstudies that assessed the association of dietary patterns, specific foods, food groups, beverages(including alcohol), macronutrients, and micronutrients with the incidence of CRC.

DATA SOURCES MEDLINE, Embase, and the Cochrane Library were searched from databaseinception to September 2019.

EVIDENCE REVIEW Only meta-analyses of prospective observational studies with a cohort studydesign were eligible. Evidence of association was graded according to established criteria as follows:convincing, highly suggestive, suggestive, weak, or not significant.

RESULTS From 9954 publications, 222 full-text articles (2.2%) were evaluated for eligibility, and 45meta-analyses (20.3%) that described 109 associations between dietary factors and CRC incidencewere selected. Overall, 35 of the 109 associations (32.1%) were nominally statistically significantusing random-effects meta-analysis models; 17 associations (15.6%) demonstrated largeheterogeneity between studies (I2 > 50%), whereas small-study effects were found for 11associations (10.1%). Excess significance bias was not detected for any association between diet andCRC. The primary analysis identified 5 (4.6%) convincing, 2 (1.8%) highly suggestive, 10 (9.2%)suggestive, and 18 (16.5%) weak associations between diet and CRC, while there was no evidence for74 (67.9%) associations. There was convincing evidence of an association of intake of red meat (highvs low) and alcohol (�4 drinks/d vs 0 or occasional drinks) with the incidence of CRC and an inverseassociation of higher vs lower intakes of dietary fiber, calcium, and yogurt with CRC risk. Theevidence for convincing associations remained robust following sensitivity analyses.

CONCLUSIONS AND RELEVANCE This umbrella review found convincing evidence of anassociation between lower CRC risk and higher intakes of dietary fiber, dietary calcium, and yogurtand lower intakes of alcohol and red meat. More research is needed on specific foods for whichevidence remains suggestive, including other dairy products, whole grains, processed meat, andspecific dietary patterns.

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341

Key PointsQuestion How credible is the evidence

behind the association of dietary factors

with colorectal cancer (CRC) risk in

published meta-analyses of prospective

observational studies?

Findings This umbrella review of 45

meta-analyses describing 109

associations found convincing evidence

for an association between lower CRC

risk and higher intakes of dietary fiber,

dietary calcium, and yogurt and lower

intakes of alcohol and red meat.

Meaning This study suggests that

dietary factors may have a role in the

development and prevention of CRC,

but more research is needed on specific

foods for which the evidence remains

suggestive.

+ Supplemental content

Author affiliations and article information arelisted at the end of this article.

Open Access. This is an open access article distributed under the terms of the CC-BY License.

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 1/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Introduction

Colorectal cancer (CRC) is the third most commonly diagnosed cancer among men and the secondmost common cancer among women worldwide.1 The etiology of CRC is multifactorial, with bothgenetic and environmental factors playing a role.2 Evidence suggests that modifiable lifestyle factors,including excess adiposity, poor diet, and physical inactivity, play an important role in the occurrenceand progression of this disease.2,3

Several systematic reviews with meta-analysis of prospective observational studies havesummarized evidence for the associations between dietary factors (eg, foods and food groups,beverages, alcohol, macronutrients, and micronutrients) and the incidence of CRC. However, to date,there has been little synthesis of the strength, precision, and quality of this evidence in aggregate.Umbrella reviews provide a structured and critical summary of the evidence and enable the gradingof evidence according to specific criteria: sample size, strength and precision of the association, andassessment of the presence of biases.4-6 The review from the World Cancer Research Fund (WCRF)in partnership with the American Institute for Cancer Research (AICR) provided a summary reportbased on systematic reviews with dose-response meta-analysis of the association between dietaryexposures and CRC. The WCRF/AICR report applied many, but not all, umbrella review criteria forevidence grading and was based on studies published until April 2015.7 To complement the WCRF/AICR report, we conducted an umbrella review of meta-analyses to provide an updated and unifiedsystematic summary of epidemiological data evaluating the strength of the overall body of evidenceinvestigating the associations between dietary factors and CRC incidence. Furthermore, we expandthe review of dietary exposures to some not previously evaluated, such as certain dietary patterns.

Methods

The protocol of this umbrella review has been registered in PROSPERO (CRD42020173636). This studyadhered to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guideline.

Search Strategy and Eligibility CriteriaWe searched MEDLINE, Embase, and the Cochrane Library from database inception to September2019 to identify meta-analyses of prospective, observational studies. The search strategy can befound in eTable 1 in the Supplement.

Two authors (Y.S.L. and T.Y.W.) independently screened titles or abstracts and examined the fulltext of potentially eligible articles. Discrepancies were resolved by a third reviewer (S.K.V.).

Studies were included that met the following criteria: (1) meta-analysis of prospectiveobservational studies (ie, cohort design) among adults with multivariable-adjusted summary riskestimates and corresponding 95% CIs that (2) investigated the association of dietary factor(s) withthe incidence of CRC. Eligible dietary factors included dietary patterns, prespecified diet qualityindices, specific foods, food groups, beverages (including alcohol), macronutrients (ie,carbohydrates, fat, protein), and micronutrients (ie, vitamins, minerals, antioxidants, polyphenols).When more than 1 meta-analysis on the same research question was available, we assessed only thestudy that included the largest data set, as previously described (eAppendix in the Supplement).4,6

We excluded meta-analyses of studies with other study designs and those with insufficient orinadequate data for quantitative synthesis.

Data ExtractionData were extracted, and the methodological quality of included meta-analyses was assessed usingA Measurement Tool to Assess Systematic Reviews (AMSTAR-2)8 by 2 authors (Y.S.L. and T.Y.W.)independently and checked by a third author (S.K.V.). For each eligible meta-analysis, we abstracteddata at the meta-analysis level (eAppendix in the Supplement). Disagreements were resolved byconsensus.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 2/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Statistical AnalysisFor each association of diet with CRC, we recalculated the adjusted summary estimates andcorresponding 95% CIs with P values using the DerSimonian and Laird random-effects model.9

Heterogeneity was assessed with the I2 statistic.10 We estimated the 95% prediction interval (PrI),which evaluates the uncertainty for the effect size that would be anticipated in a new studyaddressing the identical association.11 The evidence for small-study effects was assessed by Eggerregression asymmetry test.12 P < .10 was taken as statistical evidence of the presence of small-studyeffects. We also applied the excess significance test, which evaluates whether the observed numberof studies with statistically significant results (positive studies, P �.05) differs from the expectednumber of positive studies, by using a χ2 test.13 Excess significance bias was set at P � .10. Thestatistical analysis and power calculations were conducted using Stata version 15.0 (StataCorp).

Evaluation of the Quality of EvidenceWe graded the quality of the evidence per association generated by a meta-analysis by applyingseveral criteria in concordance with previously published umbrella reviews.4-6 In brief, associationsthat presented nominally significant random-effects summary effect sizes (ie, P � .05) were gradedas convincing (class I), highly suggestive (class II), suggestive (class III), or weak (class IV) evidence(Table 1).

Sensitivity AnalysesFor each meta-analysis initially graded as showing convincing, highly suggestive, or suggestiveevidence, we reexamined the list of adjusted confounding factors at the component study level. Weperformed a sensitivity analysis by including only adjusted estimates to assess the robustness of themain analysis. Based on the literature, the most consistent potential confounders considered forsensitivity analysis are age, body mass index (BMI), race, sex, and other dietary factors as observedin our umbrella review. Other sensitivity analyses included the omission of small-sized studies (ie,<25th percentile)14 from those meta-analyses with evidence of small-study effects, as detailed underthe heading of statistical analyses and low-quality studies (as defined in each meta-analysis).

Results

In total, we identified 9954 publications, evaluated 222 full-text articles (2.2%), and included 45meta-analyses (20.3%)15-59 describing 109 associations in this umbrella review (eFigure in theSupplement). The 177 articles (79.7%) that were excluded and the reasons for their exclusion areprovided in eTable 2 in the Supplement. The descriptive characteristics of the 45 eligiblemeta-analyses15-59 are provided in eTable 3 in the Supplement. The included meta-analyses were

Table 1. Criteria for Quality of Evidence Classificationin Observational Studiesa

Category CriteriaConvincing, class 1 • No. of cases >1000

• P < 1 × 10−6

• I2 < 50%• 95% prediction interval excluding the null• No small-study effects• No excess significance bias

Highly suggestive, class II • No. of cases >1,000• P < 1 × 10−6

• Largest study with a statistically significanteffect

Suggestive, class III • No. of cases >1,000• P < 1 × 10−3

Weak, class IV • P < .05

Nonsignificant • P > .05a Criteria in concordance with previously published umbrella reviews.4-6

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 3/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

published between 2006 and 2019. The median (interquartile range [IQR]) number of studies permeta-analysis was 6 (3-9), and the median (IQR) duration of follow-up was 10.2 (9.3-12.9) years. Themedian (IQR) meta-analysis sample size was 598 744 (229 046-991 476). The median (IQR) numberof cases (ie, incidence of CRC) was 5076 (2673-9355) cases, and the number of cases was greaterthan 1000 for 99 associations (90.8%).

The evaluation of methodological quality using AMSTAR-2 (Table 2 and Table 3; eTable 3 in theSupplement) revealed that 2 meta-analyses (4.4%) were of high quality and 15 meta-analyses(33.3%) were of moderate quality. Twenty meta-analyses (44.4%) were of low quality, with theremaining 8 (17.8%) rated as having critically low quality.

Description and Summary of AssociationsOverall, the 45 meta-analyses described 109 associations, including 794 individual study estimatesof CRC incidence associated with dietary exposures. The included meta-analyses provided adjustedsummary estimates on the associations between dietary patterns15,16,27,38,49,55 (13 associations),food groups17-24,56-59 (23 associations), beverages including alcohol25,26,28-33 (12 associations),macronutrients34-37,39,40,49,59 (18 associations), and micronutrients41-48,50-54 (43 associations) andthe incidence of CRC. Definitions of dietary patterns is provided in the eAppendix in the Supplement.

A total of 35 of the 109 associations (32.1%) were nominally statistically significant at P � .05(Table 2 and Table 3). Of these, only 7 associations (20.0%) reached statistical significance atP � 1 × 10−6. Overall, 24 significant associations (68.6%) suggested potential protective effects ofdietary factors or dietary patterns associated with CRC risk, including adherence to a healthy diet,Mediterranean diet, pesco-vegetarian diet, or semivegetarian diet and higher intakes of dietary fiber,whole grains, legumes, dairy products including yogurt and nonfermented milk, fruits andvegetables, and micronutrients (ie, supplemental and dietary calcium, zinc, magnesium, vitamin A,vitamin B6, folic acid, vitamin D, vitamin E). The remaining significant associations (31.4%) suggestedhigher risk of CRC with adherence to an unhealthy diet or Western diet and increased intake ofalcohol, red meat, processed meat, pork, eggs, and haem iron.

Seventeen associations (15.6%) had large heterogeneity (I2 > 50%). The 95% PrIs excluded thenull value for 8 associations (7.3%). The effect sizes of the largest study were statistically significantat P � .05 for 29 associations (26.6%). Small-study effects were found for 11 associations (10.1%),and excess significance bias was not identified. Seventy-four (67.9%) associations without statisticalsignificance at P � .05 using random-effects models are presented in eTable 4 in the Supplement.

Main Analysis GradingConvincing EvidenceAmong the 109 associations, 5 (4.6%) were supported by convincing evidence (Table 2 and Table 3).Two of these associations (40.0%), ie, higher vs lower red meat intake (AMSTAR-2, high quality) andheavy alcohol intake (defined as >4 drinks per day compared with those who did not drink oroccasionally drank) (AMSTAR-2, moderate quality), were associated with increased risk of CRC. Incontrast, convincing evidence was found for 3 inverse associations: higher vs lower intake of totaldietary fiber (AMSTAR-2, high quality), calcium (AMSTAR-2, moderate quality), and yogurt(AMSTAR-2, moderate quality) were associated with reduced CRC incidence.

Highly Suggestive EvidenceTwo associations (1.8%) had highly suggestive evidence of an association between diet and incidenceof CRC (Table 2 and Table 3). Higher intake of total dairy products (eg, milk, cheese, yogurt)(AMSTAR-2, high quality) was associated with significant CRC risk reduction compared with lowerintake. However, a moderate intake of alcohol (defined as >1-3 drinks but not more than 4 per day)(AMSTAR-2, moderate quality) was associated with an increase in the incidence of CRC comparedwith 0 drinks or occasionally drinking.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 4/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Tabl

e2.

Qua

lity

ofEv

iden

ceof

Asso

ciat

ions

Betw

een

the

Role

ofD

ieta

ndIn

crea

sein

Colo

rect

alCa

ncer

Inci

denc

e

Clas

sific

atio

nEx

posu

reSo

urce

Com

paris

onSu

mm

ary

met

ric

Cred

ibili

tyas

sess

men

t

AMST

AR-2

Rand

omef

fect

size

(95%

CI)

Pva

lue

I2,%

Larg

est

stud

y95

%CI

Pred

ictio

nin

terv

al95

%CI

Egge

rP

valu

e

Exce

sssi

gnifi

canc

ete

stQ

ualit

yof

evid

ence

clas

sO

/EP

valu

eFo

odRe

dm

eat

Schw

ings

hack

let

al,5

620

18H

igh

vslo

wRR

1.13

(1.0

8-1.

19)

<1×

10−

620

.51.

15-1

.19

1.02

-1.2

6.1

83/

6.0

NP

IH

igh

Alco

holic

beve

rage

Alco

hol

Fedi

rko

etal

,31

2011

≥4dr

inks

/dvs

non-

/occ

asio

nal

drin

kers

RR1.

58(1

.38-

1.80

)<1

×10

−6

0.0

1.27

-2.1

61.

33-1

.87

.80

5/5.

8N

PI

Mod

erat

e

Alco

holic

beve

rage

Alco

hol

Fedi

rko

etal

,31

2011

>1-3

drin

ks/d

vsno

n-/o

ccas

iona

ldr

inke

rs

RR1.

24(1

.14-

1.34

)<1

×10

−6

49.3

1.01

-1.1

30.

95-1

.61

<.00

19/

2.8

.77

IIM

oder

ate

Food

Proc

esse

dm

eat

Schw

ings

hack

let

al,5

620

18H

igh

vslo

wRR

1.14

(1.0

7-1.

23)

.000

125

.91.

09-1

.32

0.97

-1.3

5.9

84/

6.9

NP

IIIH

igh

Diet

ary

beha

vior

Adhe

renc

eto

Wes

tern

diet

Feng

etal

,16

2017

Hig

hvs

low

OR

1.28

(1.1

3-1.

45)

.000

172

.21.

09-1

.44

0.79

-2.0

7.1

78/

6.5

>.99

IIIM

oder

ate

Food

Eggs

Schw

ings

hack

let

al,5

620

18H

igh

vslo

wRR

1.36

(1.1

0-1.

68)

.004

0.0

1.10

-1.7

80.

35-5

.31

.28

1/1.

2N

PIV

Hig

h

Diet

ary

beha

vior

Adhe

renc

eto

alco

hol

drin

king

Feng

etal

,16

2017

Hig

hvs

low

OR

1.53

(1.0

4-2.

25)

.03

93.5

1.54

-2.1

00.

37-6

.34

.56

3/8.

3N

PIV

Mod

erat

e

Alco

holic

beve

rage

Beer

Zhan

get

al,3

2

2015

Drin

kers

vsno

n-/

occa

sion

aldr

inke

rsRR

1.08

(1.0

2-1.

15)

.01

0.0

1.04

-1.2

81.

01-1

.16

.52

1/1.

5N

PIV

Mod

erat

e

Diet

ary

beha

vior

Adhe

renc

eto

unhe

alth

ydi

eta

Gros

soet

al,2

7

2017

Hig

hvs

low

RR1.

13(1

.03-

1.23

).0

0730

.51.

02-1

.36

0.89

-1.4

3.7

62/

3.7

NP

IVLo

w

Food

Pork

Carr

etal

,57

2016

Hig

hvs

low

RR1.

17(1

.04-

1.31

).0

10.

00.

93-1

.38

0.90

-1.5

1.8

60/

0.7

NP

IVLo

w

Mic

ronu

trie

ntH

eme

iron

Qia

oet

al,5

1

2013

Hig

hvs

low

RR1.

12(1

.01-

1.24

).0

322

.60.

99-1

.29

0.90

-1.3

9.8

11/

NA

NA

IVCr

itica

llylo

w

Abbr

evia

tions

:AM

STAR

-2,A

Mea

sure

men

tToo

lto

Asse

ssSy

stem

atic

Revi

ews;

NA,

nota

pplic

able

beca

use

ofno

nsig

nific

ante

ffect

estim

ate;

NP,

notp

ertin

entb

ecau

sees

timat

ednu

mbe

risl

arge

rtha

nob

serv

edan

dth

ere

isno

evid

ence

ofex

cess

signi

fican

ceba

sed

onas

sum

ptio

nm

ade

forp

laus

ible

effe

ctsiz

e;O

/E,o

bser

ved/

expe

cted

num

bero

fstu

dies

with

signi

fican

tres

ults

;OR,

odds

ratio

;RR,

risk

ratio

.

aU

nhea

lthy

diet

was

defin

edby

auth

orsa

scha

ract

eriz

edby

,but

notl

imite

dto

,red

and

proc

esse

dm

eat,

suga

rydr

inks

and

salty

snac

ks,s

tarc

hyfo

ods,

and

refin

edca

rboh

ydra

tes.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 5/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Tabl

e3.

Qua

lity

ofEv

iden

ceof

Asso

ciat

ions

Betw

een

the

Role

ofD

ieta

ndRe

duct

ion

inCo

lore

ctal

Canc

erIn

cide

nce

Clas

sific

atio

nEx

posu

reSo

urce

Com

paris

onSu

mm

ary

met

ric

Cred

ibili

tyas

sess

men

t

AMST

AR-2

Rand

omef

fect

size

(95%

CI)

Pva

lue

I2,%

Larg

ests

tudy

95%

CIPr

edic

tion

inte

rval

95%

CIEg

ger

Pva

lue

Exce

sssi

gnifi

canc

ete

stQ

ualit

yof

evid

ence

clas

sO

/EP

valu

eM

acro

nutr

ient

Tota

ldie

tary

fiber

Reyn

olds

etal

,49

2019

Hig

hvs

low

RR0.

84(0

.78-

0.89

)<1

×10

−6

18.1

0.65

-0.8

50.

72-0

.97

.53

6/12

.1N

PI

Hig

h

Mic

ronu

trie

ntDi

etar

yca

lciu

mM

eng

etal

,50

2019

Hig

hvs

low

HR

0.77

(0.7

3-0.

82)

<1×

10−

60.

00.

75-0

.94

0.72

-0.8

3.6

05/

3.9

>.99

IM

oder

ate

Food

Yogu

rtZh

ang

etal

,24

2019

Hig

hvs

low

OR

0.81

(0.7

6-0.

86)

<1×

10−

60.

00.

75-0

.87

0.72

-0.9

0.8

42/

1.8

>.99

IM

oder

ate

Food

Dair

ypr

oduc

tsSc

hwin

gsha

ckl

etal

,56

2018

Hig

hvs

low

RR0.

83(0

.76-

0.89

)<1

×10

−6

60.3

0.83

-0.9

50.

65-1

.04

.17

8/4.

0.9

9II

Hig

h

Food

Who

legr

ains

Schw

ings

hack

let

al,5

620

18H

igh

vslo

wRR

0.88

(0.8

3-0.

94)

.000

0634

.90.

88-0

.99

0.77

-1.0

1.0

74/

1.0

.26

IIIH

igh

Diet

ary

beha

vior

Adhe

renc

eto

heal

thy

diet

aFe

nget

al,1

620

17H

igh

vslo

wO

R0.

84(0

.76-

0.92

).0

003

56.2

0.69

-0.9

00.

60-1

.17

.60

5/7.

5N

PIII

Mod

erat

e

Beve

rage

Non

ferm

ente

dm

ilkRa

lsto

net

al,3

0

2014

Hig

hvs

low

RR0.

85(0

.78-

0.93

).0

005

0.0

0.78

-1.1

80.

77-0

.94

.96

3/0.

9.9

6III

Mod

erat

e

Diet

ary

beha

vior

Pesc

o-ve

geta

rian

diet

Godo

seta

l,38

2016

Yesv

sno

RR0.

67(0

.53-

0.83

).0

004

0.0

0.48

-0.9

40.

15-2

.89

.44

2/1.

7.9

4III

Low

Diet

ary

beha

vior

Sem

iveg

etar

ian

diet

Godo

seta

l,38

2016

Yesv

sno

RR0.

86(0

.79-

0.94

).0

007

0.0

0.76

-0.9

50.

72-1

.04

.96

1/1.

2N

PIII

Low

Mic

ronu

trie

ntSu

pple

men

tal

calc

ium

Hei

ne-B

rörin

get

al,4

720

15Ye

svsn

oRR

0.88

(0.8

2-0.

94)

.000

0951

.70.

88-1

.05

0.70

-1.0

9.0

73/

3.4

NP

IIILo

w

Mic

ronu

trie

ntSu

pple

men

tal

calc

ium

Hei

ne-B

rörin

get

al,4

720

15H

igh

vslo

wRR

0.80

(0.7

2-0.

89)

.000

0230

.90.

72-1

.02

0.63

-1.0

1.8

84/

2.7

.95

IIILo

w

Diet

ary

beha

vior

Adhe

renc

eto

Med

iterr

anea

ndi

etSc

hwin

gsha

ckl

etal

,15

2017

Hig

hvs

low

RR0.

86(0

.80-

0.92

)8.

4×

10−

629

.70.

80-0

.99

0.74

-1.0

0.8

43/

5.0

NP

IIICr

itica

llylo

wFo

odVe

geta

bles

Schw

ings

hack

let

al,5

620

18H

igh

vslo

wRR

0.96

(0.9

2-1.

00)

.04

25.0

0.96

-1.0

40.

87-1

.05

.26

1/0.

9>.

99IV

Hig

h

Food

Frui

tsSc

hwin

gsha

ckl

etal

,56

2018

Hig

hvs

low

RR0.

93(0

.87-

0.99

).0

151

.80.

96-1

.06

0.78

-1.1

0.1

63/

0.9

.99

IVH

igh

Mic

ronu

trie

ntZi

ncLi

etal

,53

2014

Hig

hvs

low

RR0.

82(0

.69-

0.96

).0

267

.00.

68-0

.78

0.48

-1.3

8.4

32/

3.8

NP

IVM

oder

ate

Food

Legu

mes

Zhu

etal

,21

2015

Hig

hvs

low

RR0.

90(0

.83-

0.98

).0

146

.10.

83-1

.09

0.66

-1.2

3.1

35/

1.0

.16

IVLo

w

Mac

ronu

trie

ntCe

real

fiber

Aune

etal

,40

2011

Hig

hvs

low

RR0.

90(0

.83-

0.96

).0

030.

00.

76-0

.98

0.82

-0.9

8.3

61/

2.9

NP

IVLo

w

Mic

ronu

trie

ntM

ultiv

itam

inH

eine

-Brö

ring

etal

,47

2015

Yesv

sno

RR0.

91(0

.86-

0.97

).0

0213

.70.

88-1

.11

0.81

-1.0

2.5

52/

0.8

.91

IVLo

w

Mic

ronu

trie

ntVi

tam

inA

Hei

ne-B

rörin

get

al,4

720

15Ye

svsn

oRR

0.76

(0.6

2-0.

94)

.01

0.0

0.50

-1.1

00.

48-1

.21

.78

1/1.

7N

PIV

Low

Mic

ronu

trie

ntVi

tam

inB 6

Liu

etal

,48

2015

Hig

hvs

low

RR0.

88(0

.79-

0.99

).0

347

.20.

66-0

.99

0.62

-1.2

6.7

05/

7.1

NP

IVLo

w

Mic

ronu

trie

ntFo

licac

idLi

uet

al,4

820

15H

igh

vslo

wRR

0.88

(0.8

1-0.

95)

.002

42.9

0.58

-0.8

40.

68-1

.15

.08

7/13

.7N

PIV

Low

Mic

ronu

trie

ntVi

tam

inD

Liu

etal

,48

2015

Hig

hvs

low

RR0.

85(0

.75-

0.97

).0

144

.10.

83-1

.33

0.57

-1.2

6.4

62/

1.5

>.99

IVLo

w

Mic

ronu

trie

ntVi

tam

inE

Hei

ne-B

rörin

get

al,4

720

15Ye

svsn

oRR

0.87

(0.7

5-1.

00)

.048

0.0

0.64

-1.3

60.

73-1

.03

.75

1/N

AN

AIV

Low

Mic

ronu

trie

ntM

agne

sium

Chen

etal

,52

2012

Hig

hvs

low

RR0.

87(0

.77-

0.99

).0

30.

50.

62-1

.03

0.75

-1.0

2.6

80/

3.5

NP

IVCr

itica

llylo

w

Abbr

evia

tions

:AM

STAR

-2,A

Mea

sure

men

tToo

lto

Asse

ssSy

stem

atic

Revi

ews;

HR,

haza

rdra

tio;N

A,no

tap

plic

able

beca

use

ofno

nsig

nific

ante

ffect

estim

ate;

NP,

notp

ertin

entb

ecau

sees

timat

ednu

mbe

risl

arge

rtha

nob

serv

edan

dth

ere

isno

evid

ence

ofex

cess

signi

fican

ceba

sed

onas

sum

ptio

nm

ade

forp

laus

ible

effe

ctsiz

e;O

/E,

obse

rved

/exp

ecte

dnu

mbe

rofs

tudi

esw

ithsig

nific

antr

esul

ts;O

R,od

dsra

tio;R

R,ris

kra

tio.

aH

ealth

ydi

etw

asde

fined

byth

eau

thor

sash

igh

inta

keso

fveg

etab

les,

frui

ts,w

hole

grai

ns,o

live

oil,

fish,

soy,

poul

try,

and

low

-fatd

airy

.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 6/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Suggestive, Weak, and No EvidenceSuggestive evidence for the association of diet or dietary patterns with reduced risk of CRC wasfound for 8 associations (7.3%), including adherence to the Mediterranean diet, adherence to ahealthy diet, adherence to a pesco-vegetarian diet, adherence to a semivegetarian diet, and intakeof whole grains, nonfermented milk, and supplemental calcium (Table 3). However, there wassuggestive evidence that 2 exposures (adherence to the Western diet and intake of processed meat)were associated with increased risk of CRC in adults (Table 2). The remaining associations, with eitherweak evidence (18 [16.5%]) or no evidence (74 [67.9%]) are provided in Table 2, Table 3, and eTable 4in the Supplement.

Sensitivity AnalysesResults from sensitivity analyses are reported in eTable 5 in the Supplement. Among convincingassociations for increased risk of CRC, only alcohol exposure showed evidence for small-studyeffects. Removal of small studies from the alcohol analysis did not modify the evidence rating. Whenexcluding low-quality studies and those that did not adjust for important potential confounders, boththe association of alcohol intake and red meat intake with CRC retained their evidence ratings. Forinverse associations, all retained the class I rank following sensitivity analyses. A summary of allsensitivity analyses is presented in Table 4.

Discussion

We included 45 published meta-analyses, which comprised 109 adjusted summary risk estimates forthe associations of dietary factors with CRC incidence. We found that few of the 35 statisticallysignificant associations (ie, positive associations of 4 drinks/d and red meat with the incidence of CRCand inverse associations of higher intake of dietary fiber, calcium, and yogurt with the incidence ofCRC) were supported by convincing evidence in the main and sensitivity analyses. Suggestiveevidence exists for an inverse association of a healthy dietary pattern, Mediterranean diet, pesco-vegetarian diet, and semivegetarian diet and intake of whole grains, total dairy products, andsupplemental calcium with CRC incidence. Suggestive evidence also exists for positive associationsbetween higher intakes of processed meat and a moderate intake of alcohol (>1-3 drinks/d) and theincidence of CRC.

The continuous update project (CUP) report by the WCRF in partnership with the AICRconducted a comprehensive review including meta-analyses to draw conclusions about the roles ofdiet, nutrition, and physical activity in cancer prevention and survival.60 Although the WCRF/AICRreport provides rigorous estimates for the association of diet with CRC incidence, their criteria forstudy inclusion and grading of evidence differed from the current umbrella review (eTable 6 in theSupplement).7 The current umbrella review widened the scope of dietary exposure evaluation toinclude dietary patterns, in particular, providing evidence to support following a Mediterranean dietor overall dietary patterns that exclude red meat (ie, semivegetarian and pesco-vegetarian diets). Inaddition, yogurt, which was not evaluated as a discrete exposure in the WCRF/AICR CUP report, wasidentified as being associated with reduced risk of CRC.

Our findings largely support existing cancer prevention dietary guidance regarding increasingconsumption of dietary fiber and dairy products and limiting intake of red meat and alcoholicbeverages.7,61-63 However, we did not find convincing evidence to support limiting consumption ofprocessed meat for CRC prevention. Our results confirmed the positive association of processedmeat products with CRC; however, its credibility was suggestive. We did not measure dose-responseassociations owing to data limitations. However, we examined the summary effect size estimate forconcordance regarding the direction of association and level of statistical significance to othereligible meta-analyses reporting the same association (eTable 7 in the Supplement). Findings wereconcordant with our primary analysis regarding magnitude and direction of the association.Nonetheless, proposed biological mechanisms are highly plausible and the associations have been

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 7/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

consistent across studies, with coherence between findings in humans and preclinical studies,64

providing support for a causal relationship. It has to be noted that the criteria in this article did notexplicitly account for biologic plausibility as part of the determination for association.

Although we found convincing evidence that a higher intake of dietary fiber was associated withreduced risk of CRC, the evidence for different sources of fiber remains low. We observed asuggestive inverse association between whole grains and CRC incidence. Fiber derived from wholegrains is accompanied by micronutrients, such as folate, that have demonstrated associations withCRC, but less evidence supports a role in carcinogenesis for other fiber sources (eg, vegetables, fruit,and legumes). We observed a weak inverse association between fruit and vegetable consumptionand CRC incidence. Fruit and vegetables are abundant sources not only of fiber, but also of micro- andmacronutrients with antitumor properties; thus, they are plausible targets for dietary prevention.65

Although plausible and consistent, the associations of fruit and vegetable intake with CRC were alsoreported as weak in the WCRF/AICR reports,7,60 which noted differences in association according tosex as well as nonlinear associations that may attenuate summary estimates.

Table 4. Summary of Sensitivity Analyses

Exposure Comparison

Quality of evidence, class

CES/AMSTAR-2Primary analysis

Sensitivity analysesIncluding studiesadjusted forpotentialconfoundingvariables

Omission of small-sized studies

Omission oflow-quality studies

Dietary behaviors

Adherence to Mediterranean diet inSchwingshackl et al,15 2017

High vs low III III NA IIIa III/Critically low

Adherence to Western diet in Fenget al,16 2017

High vs low III III NA III IV/Moderate

Adherence to healthy diet in Fenget al,16 2017

High vs low III III NA III IV/Moderate

Pesco-vegetarian diet in Godoset al,38 2016

Yes vs no III III NA IIIb III/Low

Semivegetarian diet in Godoset al,38 2016

Yes vs no III III NA IIIb III/Low

Food

Red meat in Schwingshacklet al,56 2018

High vs low I I NA Ia I/High

Processed meat in Schwingshacklet al,56 2018

High vs low III III NA IIIa III/High

Whole grains in Schwingshacklet al,56 2018

High vs low III III III IIIa III/High

Dairy products in Schwingshacklet al,56 2018

High vs low II II NA IIa III/High

Yogurt in Zhang et al,24 2019 High vs low I I NA Ia I/Moderate

Beverages

Moderate alcohol in Fedirkoet al,31 2011

>1-3 drinks/d vs non-/occasional drinkers

II II III III III/Moderate

Heavy alcohol in Fedirkoet al,31 2011

≥4 drinks/d vs non-/occasional drinkers

I I NA I I/Moderate

Nonfermented milk in Ralstonet al,30 2014

High vs low III III NA III IV/Moderate

Micronutrients

Total dietary fiber in Reynoldset al,49 2019

High vs low I I NA I I/High

Dietary calcium in Menget al,50 2019

High vs low I I NA Ib I/Moderate

Supplemental calcium inHeine-Bröring et al,47 2015

Yes vs no III III III IIIa III/Low

Supplemental calcium inHeine-Bröring et al,47 2015

High vs low III III NA IIIa III/Low

Abbreviations: AMSTAR-2, A Measurement Tool to Assess Systematic Reviews; CES, classof evidence after sensitivity analyses; NA, not applicable because sensitivity analysis wasnot performed because of no evidence of small-study effects.

a No information on quality assessment of primary studies.b Meta-analysis reported all good-quality studies.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 8/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Our findings convincingly highlight dietary calcium and yogurt as associated with reduced riskof CRC incidence. Findings for cheese and milk were null and weak, respectively. We also foundsuggestive evidence for lower CRC risk with higher intakes of supplemental calcium. These findingsare complementary given that all dietary factors are sources of calcium, which binds to unconjugatedbile acids and free fatty acids in the colonic lumen to minimize their toxic effects.66 Whether dietarycalcium is a causal factor in CRC prevention is difficult to determine given that other correlatedconstituents (eg, fortified vitamin D) could account for the observations.67 The high fat content ofcheese and cream may hinder their protective associations, possibly by increasing bile acid and fattyacid excretion in the colonic lumen. Other nutrients or bioactive compounds in dairy products, suchas lactoferrin or generation of butyrate, may also play a role.67 The biologic plausibility of yogurt’sassociations with reduced risk of CRC stems from the presence of lactic acid–producing bacteria(Lactobacillus bulgaricus and Streptococcus thermophilus) that are purported to reduce levels ofcarcinogens such as nitroreductase, fecal-activated bacterial enzymes, and soluble fecal bileacids.68-70 Yogurt intake has been shown to reduce the risk of colorectal adenomas with highmalignant potential independent of calcium and nonyogurt dairy intake.71 Its protective associationcould also be mediated through its calcium content and modulated by the gut microbiome.

Selected meta-analyses provide suggestive evidence that supports the adoption of an overallhealthy dietary pattern, Mediterranean diet, pesco-vegetarian diet, and semivegetarian diet toprevent CRC, in contrast with Western dietary patterns that are associated with increased CRC risk.Overall dietary pattern accounts for the totality of dietary components in combination and theirsynergistic or antagonistic influence on human metabolism and disease. Prudent dietary patterns,characterized by higher intakes of vegetables, fruit, whole grains, and low-fat dairy products andlower intakes of alcohol and meat products, are often accompanied by a healthier lifestyle; theconverse is true for Western-type dietary patterns. Hence, the suggestive association observed forthese dietary behaviors could be influenced by the totality of a healthy or unhealthy lifestyle, inaddition to combination(s) of protective or harmful dietary factors.

LimitationsThis study has limitations. A possible limitation of our review was exclusion of dose-response meta-analyses because the data needed for predictive interval estimation and assessment of small studyand excess significant bias effects were not available in those articles. Randomized clinical trials,which account for confounding by design, are scarce in research on associations between diet andcancer owing to cost, the long follow-up time required for cancer end points, and ethical concerns.Consequently, we restricted the current umbrella review to meta-analyses of prospectiveobservational studies. Some forms of bias, such as recall bias owing to self-reported diet, are possiblebut likely to be nondifferential, which would attenuate the observed associations. We generatedestimates of publication bias by assessing small-study effects using the Egger test. However, theEgger test is not recommended with the inclusion of less than 10 studies.72 Although the small-studyeffect was indicated for only 11 meta-analyses (10.1%), among these 4 (36.4%) included 5 to 10studies and 2 (18.2%) included less than 5 studies. However, we were unable to perform alternatetests, such as the Peters test, because this required cases and noncases for each level of exposure,and this information was largely lacking in the meta-analyses.73 Thus, more research is needed toinvestigate the associations based on small numbers of included studies. A further limitation is thatwe did not conduct subgroup analysis (eg, by sex, age group, or location of cancer, such as colon orrectum) because of the lack of data for grading the quality of evidence for most of the exposures.Dietary associations with CRC may differ according to sex and tumor location, as was highlighted bythe WCRF/AICR reports.7,60

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 9/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

Conclusions

The findings of this study support existing recommendations for diet in the primary prevention ofCRC, emphasizing higher intakes of dietary fiber, calcium, and yogurt and lower intakes of red meatand alcohol. Emerging evidence supports a possible role for overall dietary patterns that, in totality,emphasize habitually consuming fruits, vegetables, grains, and low-fat dairy and reducing red meatand alcohol intake. More research is needed on specific foods for which evidence remains suggestive,including other dairy products, whole grains, processed meat, and specific dietary patterns.

ARTICLE INFORMATIONAccepted for Publication: December 23, 2020.

Published: February 16, 2021. doi:10.1001/jamanetworkopen.2020.37341

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Veettil SK et al.JAMA Network Open.

Corresponding Author: Nathorn Chaiyakunapruk, Pharm D, PhD, Department of Pharmacotherapy, College ofPharmacy, The University of Utah, 30 2000 E, Salt Lake City, UT 84112 (nathorn.chaiyakunapruk@utah.edu).

Author Affiliations: Department of Pharmacotherapy, College of Pharmacy, University of Utah, Salt Lake City(Veettil, Chaiyakunapruk); School of Pharmacy, International Medical University, Bukit Jalil, Kuala Lumpur, Malaysia(Wong, Loo); Department of Nutrition and Integrative Physiology, University of Utah, Salt Lake City (Playdon);Huntsman Cancer Institute, Cancer Control and Population Sciences Program, University of Utah, Salt Lake City(Playdon); School of Medicine, Faculty of Health and Medical Sciences, Taylor’s University, Subang Jaya, Selangor,Malaysia (Lai); School of Pharmacy, Monash University Malaysia, Bandar Sunway, Subang Jaya, Selangor, Malaysia(Lai); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts(Giovannucci); Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts(Giovannucci).

Author Contributions: Drs Veettil and Chaiyakunapruk had full access to all of the data in the study and takeresponsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Veettil, Wong, Loo, Giovannucci, Chaiyakunapruk.

Acquisition, analysis, or interpretation of data: Veettil, Wong, Loo, Playdon, Lai, Giovannucci.

Drafting of the manuscript: Veettil, Wong, Loo, Giovannucci.

Critical revision of the manuscript for important intellectual content: Veettil, Playdon, Lai, Giovannucci,Chaiyakunapruk.

Statistical analysis: Veettil, Wong, Loo, Chaiyakunapruk.

Administrative, technical, or material support: Veettil, Chaiyakunapruk.

Supervision: Chaiyakunapruk.

Conflict of Interest Disclosures: None reported.

Additional Contributions: The authors wish to thank George Markozannes, PhD, from the University of IoanninaSchool of Medicine, for his advice during the statistical analysis of this project. No compensation was received forhis role.

Additional Information: Data were extracted from published meta-analyses, all of which are available andaccessible.

REFERENCES1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCANestimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394-424. doi:10.3322/caac.21492

2. Chan AT, Giovannucci EL. Primary prevention of colorectal cancer. Gastroenterology. 2010;138(6):2029-2043.e10. doi:10.1053/j.gastro.2010.01.057

3. Wei EK, Giovannucci E, Wu K, et al. Comparison of risk factors for colon and rectal cancer. Int J Cancer. 2004;108(3):433-442. doi:10.1002/ijc.11540

4. Dragioti E, Solmi M, Favaro A, et al. Association of antidepressant use with adverse health outcomes:a systematic umbrella review. JAMA Psychiatry. 2019;76(12):1241-1255. doi:10.1001/jamapsychiatry.2019.2859

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 10/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

5. Li X, Meng X, Timofeeva M, et al. Serum uric acid levels and multiple health outcomes: umbrella review ofevidence from observational studies, randomised controlled trials, and Mendelian randomisation studies. BMJ.2017;357:j2376. doi:10.1136/bmj.j2376

6. Theodoratou E, Tzoulaki I, Zgaga L, Ioannidis JPA. Vitamin D and multiple health outcomes: umbrella review ofsystematic reviews and meta-analyses of observational studies and randomised trials. BMJ. 2014;348:g2035.doi:10.1136/bmj.g2035

7. World Cancer Research Fund/American Institute for Cancer Research. Diet, nutrition, physical activity andcolorectal cancer. Accessed September 13, 2020. https://www.wcrf.org/sites/default/files/Colorectal-cancer-report.pdf

8. Shea BJ, Reeves BC, Wells G, et al. AMSTAR 2: a critical appraisal tool for systematic reviews that includerandomised or non-randomised studies of healthcare interventions, or both. BMJ. 2017;358:j4008. doi:10.1136/bmj.j4008

9. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177-188. doi:10.1016/0197-2456(86)90046-2

10. Riley RD, Higgins JPT, Deeks JJ. Interpretation of random effects meta-analyses. BMJ. 2011;342(7804):d549.doi:10.1136/bmj.d549

11. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-560. doi:10.1136/bmj.327.7414.557

12. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test.BMJ. 1997;315(7109):629-634. doi:10.1136/bmj.315.7109.629

13. Ioannidis JPA, Trikalinos TA. An exploratory test for an excess of significant findings. Clin Trials. 2007;4(3):245-253. doi:10.1177/1740774507079441

14. Dechartres A, Altman DG, Trinquart L, Boutron I, Ravaud P. Association between analytic strategy andestimates of treatment outcomes in meta-analyses. JAMA. 2014;312(6):623-630. doi:10.1001/jama.2014.8166

15. Schwingshackl L, Schwedhelm C, Galbete C, Hoffmann G. Adherence to Mediterranean Diet and risk of cancer:an updated systematic review and meta-analysis. Nutrients. 2017;9(10):E1063. doi:10.3390/nu9101063

16. Feng Y-L, Shu L, Zheng P-F, et al. Dietary patterns and colorectal cancer risk: a meta-analysis. Eur J CancerPrev. 2017;26(3):201-211. doi:10.1097/CEJ.0000000000000245

17. Wu QJ, Yang Y, Vogtmann E, et al. Cruciferous vegetables intake and the risk of colorectal cancer: a meta-analysis of observational studies. Ann Oncol. 2013;24(4):1079-1087. doi:10.1093/annonc/mds601

18. Zhu B, Zou L, Qi L, Zhong R, Miao X. Allium vegetables and garlic supplements do not reduce risk of colorectalcancer, based on meta-analysis of prospective studies. Clin Gastroenterol Hepatol. 2014;12(12):1991-2001.e1. doi:10.1016/j.cgh.2014.03.019

19. Chiavarini M, Minelli L, Fabiani R. Garlic consumption and colorectal cancer risk in man: a systematic reviewand meta-analysis. Public Health Nutr. 2016;19(2):308-317. doi:10.1017/S1368980015001263

20. Turati F, Guercio V, Pelucchi C, La Vecchia C, Galeone C. Colorectal cancer and adenomatous polyps in relationto allium vegetables intake: a meta-analysis of observational studies. Mol Nutr Food Res. 2014;58(9):1907-1914.doi:10.1002/mnfr.201400169

21. Zhu B, Sun Y, Qi L, Zhong R, Miao X. Dietary legume consumption reduces risk of colorectal cancer: evidencefrom a meta-analysis of cohort studies. Sci Rep. 2015;5:8797. doi:10.1038/srep08797

22. Lu D, Pan C, Ye C, et al. Meta-analysis of soy consumption and gastrointestinal cancer risk. Sci Rep. 2017;7(1):4048. doi:10.1038/s41598-017-03692-y

23. Aune D, Lau R, Chan DSM, et al. Dairy products and colorectal cancer risk: a systematic review and meta-analysis of cohort studies. Ann Oncol. 2012;23(1):37-45. doi:10.1093/annonc/mdr269

24. Zhang K, Dai H, Liang W, Zhang L, Deng Z. Fermented dairy foods intake and risk of cancer. Int J Cancer. 2019;144(9):2099-2108. doi:10.1002/ijc.31959

25. Chen Y, Wu Y, Du M, et al. An inverse association between tea consumption and colorectal cancer risk.Oncotarget. 2017;8(23):37367-37376. doi:10.18632/oncotarget.16959

26. Wang Z-H, Gao Q-Y, Fang J-Y. Green tea and incidence of colorectal cancer: evidence from prospective cohortstudies. Nutr Cancer. 2012;64(8):1143-1152. doi:10.1080/01635581.2012.718031

27. Grosso G, Bella F, Godos J, et al. Possible role of diet in cancer: systematic review and multiple meta-analysesof dietary patterns, lifestyle factors, and cancer risk. Nutr Rev. 2017;75(6):405-419. doi:10.1093/nutrit/nux012

28. Sun C-L, Yuan J-M, Koh W-P, Yu MC. Green tea, black tea and colorectal cancer risk: a meta-analysis ofepidemiologic studies. Carcinogenesis. 2006;27(7):1301-1309. doi:10.1093/carcin/bgl024

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 11/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

29. Gan Y, Wu J, Zhang S, et al. Association of coffee consumption with risk of colorectal cancer: a meta-analysisof prospective cohort studies. Oncotarget. 2017;8(12):18699-18711. doi:10.18632/oncotarget.8627

30. Ralston RA, Truby H, Palermo CE, Walker KZ. Colorectal cancer and nonfermented milk, solid cheese, andfermented milk consumption: a systematic review and meta-analysis of prospective studies. Crit Rev Food SciNutr. 2014;54(9):1167-1179. doi:10.1080/10408398.2011.629353

31. Fedirko V, Tramacere I, Bagnardi V, et al. Alcohol drinking and colorectal cancer risk: an overall and dose-response meta-analysis of published studies. Ann Oncol. 2011;22(9):1958-1972. doi:10.1093/annonc/mdq653

32. Zhang C, Zhong M. Consumption of beer and colorectal cancer incidence: a meta-analysis of observationalstudies. Cancer Causes Control. 2015;26(4):549-560. doi:10.1007/s10552-015-0532-5

33. Xu W, Fan H, Han Z, Liu Y, Wang Y, Ge Z. Wine consumption and colorectal cancer risk: a meta-analysis ofobservational studies. Eur J Cancer Prev. 2019;28(3):151-158. doi:10.1097/CEJ.0000000000000444

34. Liu L, Zhuang W, Wang R-Q, et al. Is dietary fat associated with the risk of colorectal cancer? a meta-analysis of13 prospective cohort studies. Eur J Nutr. 2011;50(3):173-184. doi:10.1007/s00394-010-0128-5

35. Kim M, Park K. Dietary fat intake and risk of colorectal cancer: a systematic review and meta-analysis ofprospective studies. Nutrients. 2018;10(12):E1963. doi:10.3390/nu10121963

36. Chen G-C, Qin L-Q, Lu D-B, et al. N-3 polyunsaturated fatty acids intake and risk of colorectal cancer: meta-analysis of prospective studies. Cancer Causes Control. 2015;26(1):133-141. doi:10.1007/s10552-014-0492-1

37. Aune D, Chan DSM, Lau R, et al. Carbohydrates, glycemic index, glycemic load, and colorectal cancer risk:a systematic review and meta-analysis of cohort studies. Cancer Causes Control. 2012;23(4):521-535. doi:10.1007/s10552-012-9918-9

38. Godos J, Bella F, Sciacca S, Galvano F, Grosso G. Vegetarianism and breast, colorectal and prostate cancer risk:an overview and meta-analysis of cohort studies. J Hum Nutr Diet. 2017;30(3):349-359. doi:10.1111/jhn.12426

39. Lai R, Bian Z, Lin H, Ren J, Zhou H, Guo H. The association between dietary protein intake and colorectalcancer risk: a meta-analysis. World J Surg Oncol. 2017;15(1):169. doi:10.1186/s12957-017-1241-1

40. Aune D, Chan DSM, Lau R, et al. Dietary fibre, whole grains, and risk of colorectal cancer: systematic reviewand dose-response meta-analysis of prospective studies. BMJ. 2011;343:d6617. doi:10.1136/bmj.d6617

41. Bo Y, Sun J, Wang M, Ding J, Lu Q, Yuan L. Dietary flavonoid intake and the risk of digestive tract cancers:a systematic review and meta-analysis. Sci Rep. 2016;6:24836. doi:10.1038/srep24836

42. Chang H, Lei L, Zhou Y, Ye F, Zhao G. Dietary flavonoids and the risk of colorectal cancer: an updated meta-analysis of epidemiological studies. Nutrients. 2018;10(7):E950. doi:10.3390/nu10070950

43. Grosso G, Godos J, Lamuela-Raventos R, et al. A comprehensive meta-analysis on dietary flavonoid and lignanintake and cancer risk: level of evidence and limitations. Mol Nutr Food Res. 2017;61(4). doi:10.1002/mnfr.201600930

44. He X, Sun L-M. Dietary intake of flavonoid subclasses and risk of colorectal cancer: evidence from populationstudies. Oncotarget. 2016;7(18):26617-26627. doi:10.18632/oncotarget.8562

45. Jiang R, Botma A, Rudolph A, Hüsing A, Chang-Claude J. Phyto-oestrogens and colorectal cancer risk:a systematic review and dose-response meta-analysis of observational studies. Br J Nutr. 2016;116(12):2115-2128.doi:10.1017/S0007114516004360

46. Panic N, Nedovic D, Pastorino R, Boccia S, Leoncini E. Carotenoid intake from natural sources and colorectalcancer: a systematic review and meta-analysis of epidemiological studies. Eur J Cancer Prev. 2017;26(1):27-37. doi:10.1097/CEJ.0000000000000251

47. Heine-Bröring RC, Winkels RM, Renkema JMS, et al. Dietary supplement use and colorectal cancer risk:a systematic review and meta-analyses of prospective cohort studies. Int J Cancer. 2015;136(10):2388-2401. doi:10.1002/ijc.29277

48. Liu Y, Yu Q, Zhu Z, et al. Vitamin and multiple-vitamin supplement intake and incidence of colorectal cancer:a meta-analysis of cohort studies. Med Oncol. 2015;32(1):434. doi:10.1007/s12032-014-0434-5

49. Reynolds A, Mann J, Cummings J, Winter N, Mete E, Te Morenga L. Carbohydrate quality and human health:a series of systematic reviews and meta-analyses. Lancet. 2019;393(10170):434-445. doi:10.1016/S0140-6736(18)31809-9

50. Meng Y, Sun J, Yu J, Wang C, Su J. Dietary intakes of calcium, iron, magnesium, and potassium elements andthe risk of colorectal cancer: a meta-analysis. Biol Trace Elem Res. 2019;189(2):325-335. doi:10.1007/s12011-018-1474-z

51. Qiao L, Feng Y. Intakes of heme iron and zinc and colorectal cancer incidence: a meta-analysis of prospectivestudies. Cancer Causes Control. 2013;24(6):1175-1183. doi:10.1007/s10552-013-0197-x

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 12/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

52. Chen G-C, Pang Z, Liu Q-F. Magnesium intake and risk of colorectal cancer: a meta-analysis of prospectivestudies. Eur J Clin Nutr. 2012;66(11):1182-1186. doi:10.1038/ejcn.2012.135

53. Li P, Xu J, Shi Y, et al. Association between zinc intake and risk of digestive tract cancers: a systematic reviewand meta-analysis. Clin Nutr. 2014;33(3):415-420. doi:10.1016/j.clnu.2013.10.001

54. Zhou Z-Y, Wan X-Y, Cao J-W. Dietary methionine intake and risk of incident colorectal cancer: a meta-analysisof 8 prospective studies involving 431,029 participants. PLoS One. 2013;8(12):e83588. doi:10.1371/journal.pone.0083588

55. Liu Y, Tang W, Zhai L, et al. Meta-analysis: eating frequency and risk of colorectal cancer. Tumour Biol. 2014;35(4):3617-3625. doi:10.1007/s13277-013-1479-3

56. Schwingshackl L, Schwedhelm C, Hoffmann G, et al. Food groups and risk of colorectal cancer. Int J Cancer.2018;142(9):1748-1758. doi:10.1002/ijc.31198

57. Carr PR, Walter V, Brenner H, Hoffmeister M. Meat subtypes and their association with colorectal cancer:systematic review and meta-analysis. Int J Cancer. 2016;138(2):293-302. doi:10.1002/ijc.29423

58. Wu S, Feng B, Li K, et al. Fish consumption and colorectal cancer risk in humans: a systematic review and meta-analysis. Am J Med. 2012;125(6):551-9.e5. doi:10.1016/j.amjmed.2012.01.022

59. Aune D, Lau R, Chan DSM, et al. Nonlinear reduction in risk for colorectal cancer by fruit and vegetable intakebased on meta-analysis of prospective studies. Gastroenterology. 2011;141(1):106-118. doi:10.1053/j.gastro.2011.04.013

60. World Cancer Research Fund. Diet, nutrition, physical activity and cancer: a global perspective. AccessedSeptember 13, 2020. https://www.wcrf.org/dietandcancer

61. Johnston BC, Zeraatkar D, Han MA, et al. Unprocessed red meat and processed meat consumption: dietaryguideline recommendations from the Nutritional Recommendations (NutriRECS) Consortium. Ann Intern Med.2019;171(10):756-764. doi:10.7326/M19-1621

62. LoConte NK, Brewster AM, Kaur JS, Merrill JK, Alberg AJ. Alcohol and cancer: a statement of the AmericanSociety of Clinical Oncology. J Clin Oncol. 2018;36(1):83-93. doi:10.1200/JCO.2017.76.1155

63. Kim YI. AGA technical review: impact of dietary fiber on colon cancer occurrence. Gastroenterology. 2000;118(6):1235-1257. doi:10.1016/S0016-5085(00)70377-5

64. Cross AJ, Sinha R. Meat-related mutagens/carcinogens in the etiology of colorectal cancer. Environ MolMutagen. 2004;44(1):44-55. doi:10.1002/em.20030

65. Steinmetz KA, Potter JD. Vegetables, fruit, and cancer: II, mechanisms. Cancer Causes Control. 1991;2(6):427-442. doi:10.1007/BF00054304

66. Newmark HL, Wargovich MJ, Bruce WR. Colon cancer and dietary fat, phosphate, and calcium: a hypothesis.J Natl Cancer Inst. 1984;72(6):1323-1325.

67. Norat T, Riboli E. Dairy products and colorectal cancer: a review of possible mechanisms and epidemiologicalevidence. Eur J Clin Nutr. 2003;57(1):1-17. doi:10.1038/sj.ejcn.1601522

68. de Moreno de Leblanc A, Perdigón G. Yogurt feeding inhibits promotion and progression of experimentalcolorectal cancer. Med Sci Monit. 2004;10(4):BR96-BR104.

69. Tavan E, Cayuela C, Antoine J-M, Trugnan G, Chaugier C, Cassand P. Effects of dairy products on heterocyclicaromatic amine-induced rat colon carcinogenesis. Carcinogenesis. 2002;23(3):477-483. doi:10.1093/carcin/23.3.477

70. Abdelali H, Cassand P, Soussotte V, Daubeze M, Bouley C, Narbonne JF. Effect of dairy products on initiationof precursor lesions of colon cancer in rats. Nutr Cancer. 1995;24(2):121-132. doi:10.1080/01635589509514400

71. Zheng X, Wu K, Song M, et al. Yogurt consumption and risk of conventional and serrated precursors ofcolorectal cancer. Gut. 2020;69(5):970-972. doi:10.1136/gutjnl-2019-318374

72. Sterne JAC, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests andprevalence in the literature. J Clin Epidemiol. 2000;53(11):1119-1129. doi:10.1016/S0895-4356(00)00242-0

73. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication biasin meta-analysis. JAMA. 2006;295(6):676-680. doi:10.1001/jama.295.6.676

SUPPLEMENT.eAppendix. Supplementary MethodseTable 1. Search StrategyeFigure. Study Flow DiagrameTable 2. Excluded Studies

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 13/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021

eTable 3. Descriptive Characteristics of Included Meta-analyseseTable 4. Associations With Nonsignificant EvidenceeTable 5. Sensitivity Analyses for Associations With Class I, II, or III EvidenceeTable 6. Evidence Criteria: Difference Between and Comparison of WCRF and Present RevieweTable 7. Summary Estimates for Concordance in Meta-analyses: Red Meat Intake and Incidence of CRCeReferences.

JAMA Network Open | Oncology Role of Diet in Colorectal Cancer Incidence

JAMA Network Open. 2021;4(2):e2037341. doi:10.1001/jamanetworkopen.2020.37341 (Reprinted) February 16, 2021 14/14

Downloaded From: https://jamanetwork.com/ on 11/18/2021